K. Winkler et al., page 1

Electroluminescence from spatially confined exciton polaritons in a textured microcavity

K. Winkler1, C. Schneider1, J. Fischer1, A. Rahimi-Iman1, M. Amthor1, A. Forchel1, S.

Reitzenstein1,2, S. Höfling1 and M. Kamp1

1Technische Physik, Physikalisches Institut and Wilhelm Conrad Röntgen-Research Center

for Complex Material Systems, Universität Würzburg, Am Hubland, D-97074,

Würzburg, Germany

2Present address: Institut für Festkörperphysik, Technische Universität Berlin,

Hardenbergstrasse 36, 10623 Berlin, Germany

We report on the formation of spatially confined exciton-polaritons under electrical injection

in a textured microcavity. The trapping of polaritons in the diode sample is achieved through

a locally elongated GaAs microcavity with a quality factor exceeding 6000. The polaritonic

resonances of traps with diameters of 10 µm and 2 µm are studied by angular-resolved

electroluminescence spectroscopy, revealing their hybrid light-matter nature.

K. Winkler et al., page 2

Strong coupling between quantum well (QW) excitons and photons in a semiconductor

microcavity resonator leads to the formation of exciton polaritons1,2. This results in two new

eigenstates of the coupled system, namely the upper polariton (UP) and lower polariton (LP)

branch3, that are split in energy. Due to their bosonic character, the polaritons can undergo a

dynamical condensation above a critical particle density resulting in a macroscopically

occupied single particle energetic ground state4,5. This condensation is possible at elevated

temperatures because of the light effective mass of polaritons, in comparison to excitons or

atoms. Since the polaritonic ground state can acquire a high degree of spatial4,6 and temporal7–

9 coherence, such coupled light-matter systems are promising candidates for a new kind of

laser device, namely the polariton laser2.

Trapping polaritons in real space is crucial for the observation of long range coherence

phenomena, since long range order over infinite distances cannot exist in a two-dimensional

system10. While vertical trapping is already achieved by the epitaxial microcavity structure,

different approaches have been undertaken to laterally confine polaritons. They involve either

trapping the photonic11–13 or the excitonic5 part of the polariton.

One important aspect for a real device application is the formation of polaritons by means of

electrical injection instead of the commonly used optical excitation. This has been

successfully demonstrated for quasi-2D polaritons14–16. The unambiguous demonstration of a

polariton condensate under electrical pumping has so far been elusive, however it can be

expected that a carefully designed trapping potential in a polariton light emitting diode would

be highly beneficial to demonstrate this important milestone17. It has also been predicted that

spatially tightly confined polaritons of narrow emission linewidth can emit single photons on

demand18 or can be exploited in more complex schemes of scalable quantum information

processing and quantum cryptography19. This further explains the high interest in advanced

and flexible trapping schemes of microcavity exciton polaritons.

K. Winkler et al., page 3

Here we report polariton emission from an electrically driven microcavity comprising a trap

that confined the photonic part of the polariton in all three dimensions. The trap design of our

structure is based on the confinement of the photonic component through a locally elongated

cavity. In this approach, the quantum wells in the microcavity are not damaged by etching

through the active region, and detrimental surface effects are hence suppressed. This is in

stark contrast to common approaches to confine polaritons in micropillar-cavities, where

broadening of the QW emission caused by the device fabrication is usually significant and

undesired surface recombination of carriers plays a role.

The investigated sample is schematically shown in Fig. 1(a) and was grown by molecular

beam epitaxy (MBE). The intrinsic one-λ thick GaAs cavity is surrounded by two doped

GaAs/AlAs distributed Bragg reflector (DBR) mirrors grown on p-doped GaAs substrate. In

the center of the cavity a 15nm thick Ga0.88In0.12As quantum well is embedded as the active

medium. The lower/upper DBR consists of 27/23 mirror pairs and is p/n-doped by

carbon/silicon to facilitate current injection. To reduce doping related optical absorption in the

DBRs, the doping density is reduced from n = 3*1018 1/cm3 in the outer layers of the DBR to

n = 1*1018 1/cm3 near the central cavity layer (indicated in Fig. 1(a) by black to grey scale).

The cavity layer and the DBRs are designed for a low temperature resonance wavelength of

λ = 894 nm at the center of the wafer. Due to the growth-related decrease in thickness towards

the edge of the wafer, the cavity resonance can, however, be tuned down to a minimum

wavelength of λ = 857 nm.

Quantization of polaritons is achieved by processing the sample in a manner that the photonic

component of polaritons is confined in all three spatial dimensions. For this purpose, the

cavity length is locally elongated in form of circular mesas with diameters of 0.5 to 10 µm

K. Winkler et al., page 4

and a height of 5 nm. This results in a spatially energy-wise lowered cavity resonance and

therefore in a 3D-confinement of the photonic polariton component via photonic traps11.

To realize such traps, we extended the process strategy outlined before20. After growth of the

bottom DBR and the full intrinsic one-λ cavity, the sample was transferred out of the MBE

machine. Next, we performed position dependent reflection measurements at room

temperature to estimate the layer thicknesses in the bottom DBR and the cavity by comparing

the reflection spectrum with transfer-matrix calculations. If the determined cavity resonance

matches the excitonic emission energy within the radial tuning range, the full 3” wafer is

processed in order to define the optical traps prior to overgrowth with the top DBR. First, the

full wafer is patterned with alignment marks and large mesa structures via optical lithography

and wet etching21. These alignment marks serve as reference points for the subsequent

electron beam lithography (EBL) step, in which circular mesas with diameters between 500

nm and 10 µm are defined. After developing the EBL resist and the evaporation of an Al

mask, a lift-off procedure was performed, resulting in Al disks of the given diameters. The Al

disk pattern was then transferred into the semiconductor by wet etching (using a highly

diluted GaAs selective etchant), resulting in mesas of ~5 nm height, and the Al circles were

removed subsequently by dipping the sample into n-methylpyrrolidon. Prior to overgrowth of

these mesas, activated hydrogen cleaning of the surface was carried out in the load lock

chamber of the MBE to remove surface oxides and contaminations 21.

The quality factor (Q-factor) of the structure was extracted experimentally at a large Cavity-

Exciton detuning Ec- Ex < -8.02 meV by means of micro-photoluminescence (µPL)

spectroscopy. Fig. 1b) depicts a spectrum of the highly photonic fundamental mode of a 10

µm diameter trap, from which we can extract a Q-factor of ~ 6150.

The final process of electrical contacting includes dry-etching of larger pillars with a diameter

of 30 µm around the traps and planarization of the sample with isolating benzocyclobutene.

K. Winkler et al., page 5

The positioning of the large pillars relative to the buried mesas for polariton trapping is

facilitated by the alignment marks initially patterned into the structure. The contacting is

carried out by gold ring structures on top of the pillar structures (shown in Fig. 1(c)) and a

gold back-contact. A row of five different trap sizes can be electrically driven in parallel by

means of a common contact pad.

The advantages of this trap fabrication technique are the possibility to study emission from

two dimensional (planar) polaritons at the same spot as emission from the traps, and that the

traps can be two-dimensionally arranged in arbitrary configurations. This technological route

is therefore also very promising to define tailored two-dimensionally patterned potential

landscapes19,22,23 where the electrical contact could be utilized for Stark-tuning of the exciton

resonance and therewith serve as a tool for fast external manipulation.

To map out the tuning behavior and the Rabi-splitting of the polariton branches, we

performed µPL measurements in a momentum-space spectroscopy setup on traps and planar

regions by continuous-wave excitation with a CW- diode laser emitting at a wavelength of

658 nm. The sample is mounted in a helium-flow cryostat with electrical feedthroughs, at

temperatures of 5-8 K. This method allows the investigation of momentum-space dependent

energy dispersions and, thus, makes it possible to identify lower and upper polariton branches

and distinguish between trap emission and planar emission.

Fig. 2(a) shows the typical anticrossing behavior of upper and lower planar polaritons in the

region around zero detuning between cavity mode (Ec) and exciton (Ex) energy at k|| = 0

obtained from momentum resolved photoluminescence measurements as depicted exemplarily

in Fig. 2(b) for a negative detuning of -1.77 meV . In the investigated energy range, the

exciton energy can be treated as constant since it shifts significantly less (~0.2 meV) than the

K. Winkler et al., page 6

photon resonance (~ 5.5 meV) . Fitting the experimental data (solid lines in Fig. 2(a)) yields a

Rabi-splitting energy ER = 2.63 ± 0.08 meV.

The energy dispersion of a trap device with a diameter of 10 µm at a moderate red detuning

of -0.9 meV recorded under optical excitation is exemplary shown in Fig. 2(c). Due to the

large effective mass of the polaritons in this regime, the direct influence of the trapping

potential on the dispersion is not immediately visible. However, some mode quantization

already evolves close to the fundamental resonance, consistent with reports on similar

devices11,20. The visibility of the quantized modes is stronger pronounced in the red detuning

regime (Ec-Ex~ -4.6 meV): The photon confinement leads to energy modes quantized in

energy and momentum space24, in contrast to the continuous dispersion of the planar polariton

in Fig. 2(b). Because of the comparatively large trap size of 10 µm, the momentum-quantized

eigenenergies of the ground mode still resemble a two-dimensional dispersion. The

luminescence of the highly excitonic planar LP is outshined by the bright emission of the

photonic UP and the emission from the trapped LP. This is a consequence of the optical

excitation scheme with a spot of ~4 µm diameter, which predominantly generates excitons in

the region of the polariton trap. The maxima of this dispersion are fitted together with the

planar background and the results are shown in Fig. 3(b). From this evaluation, we can extract

a trapping potential of photonic fundamental mode of ΔEC = 6 meV in agreement with

transfer-matrix computations on the planar devices, and a Rabi-splitting of 2.76 meV, which

is comparable to the planar structure.

Fig. 3(c) depicts the energy momentum dispersion from an electrically contacted device at a

similar detuning (Ec-Ex~ -4.5 meV), acquired under electrical injection. The diode structure

was forward biased at a current of 0.68 mA in continuous wave mode. The emission pattern,

quality factors and the evaluation results of the electroluminescence (EL) spectra (Fig. 3(d))

are comparable to the one obtained from photoluminescense (PL) data. In EL the ground-

K. Winkler et al., page 7

mode shows a linewidth of ~330 µeV, which is in full agreement with the polariton linewidth

of ~315 µeV extracted from µPL measurements in fig. 3(a). However, the Rabi splitting is

slightly reduced to ~2.3 meV which is attributed to enhanced exciton screening effects caused

by free carriers in the electrical injection scheme at the selected excitation conditions.

We further investigated the behavior of the system under variation of the injection current for

a 2µm trap (Fig. 4) at a negative detuning of -4.2 meV. For this trap size only the fundamental

polaritonic trap mode is observable because the energy of the photonic ground mode is

increased by approximately 2 meV (compared to the 10 µm diameter traps) and the energy

splitting of the photonic resonances is predicted to be ~ 5 meV. For low injection currents

(Fig. 4(a)), the strong coupling regime directly manifests itself in the planar background

emission signal, the dispersion of which strongly deviates from a weakly coupled photonic

parabola dispersion (black line, Fig. 4a). Furthermore, the planar polariton is split into the

upper and lower branch, which are both visible at these excitation conditions. The Rabi-

splitting reduces with injection current due to exciton screening and a successive reduction of

the exciton oscillator strength25, when the system enters the weak coupling regime. This is

obvious from the high- excitation power spectrum recorded at an injection current of 4.2 mA

in Fig. 4(b), where only one dominant branch can be observed in the planar emission, which

can be approximated by the purely photonic parabolic dispersion.

The transition from strong coupling to weak coupling is accompanied by the blueshift of the

trapped polaritonic fundamental mode (Fig. 4(c)). A continuous blueshift of this mode

towards the weakly coupled photonic trap energy position (dashed blue line) is observable in

the selected pumping range. In order to assess the nonlinear properties of the system, the

integrated intensity emitted from the polariton trap is plotted as a function of the injection

current in double logarithmic style. The input-output characteristics of the device can be

K. Winkler et al., page 8

approximated by a super-linear power dependence with I ~ J1.6, where I is the integrated

intensity and J denotes the injection current through the device. This distinct power law

behavior is observed over almost two orders of magnitude of intensity, from the clearly

polaritonic regime into the transition regime from strong to weak coupling. We attribute the

deviation from the expected linear input-output behavior to an enhancement of polariton-

phonon scattering towards the energetic polariton goundstate: The loss of translational

invariance caused by the spatial confinement results in an enhanced overlap of adjacent

polariton wavefunctions in momentum space, and hence the stringent conditions on phonon k-

vectors involved in the scattering process is lifted17.

In conclusion, we demonstrated emission from trapped exciton-polaritons under electrical

injection. The trap-confinement is realized by a locally elongated, circular cavity, which can

be conveniently contacted to facilitate electrical pumping and manipulation. The polaritonic

character of the emission is confirmed both in the emission from the trap states, as well as the

planar background emission under optical and electrical pumping. Enhanced scattering

towards the groundstate in such a low-dimensional system is identified by its nonlinear input-

output characteristics, pointing towards the possibility to observe polariton condensation

under electrical injection in a sample with a larger number of quantum wells.

This work has been supported by the State of Bavaria. Technical assistance by T. Steinl and

M. Emmerling during sample preparation is gratefully acknowledged. We thank V.D.

Kulakovskii for fruitful discussions.

References

1 C. Weisbuch, M. Nishioka, A. Ishikawa, and Y. Arakawa, Physical Review Letters 69, 3314–3317 (1992).

K. Winkler et al., page 9

2 A. Kavokin, J.J. Baumberg, G. Malpuech, and F.P. Laussy, Microcavities (Oxford University Press, 2007).

3 T.C.H.C.H. Liew, I. a. A. Shelykh, and G. Malpuech, Physica E: Low-dimensional Systems and Nanostructures 43, 1543–1568 (2011).

4 J. Kasprzak, M. Richard, S. Kundermann, A. Baas, P. Jeambrun, J.M.J. Keeling, F.M. Marchetti, M.H. Szymańska, R. André, J.L. Staehli, V. Savona, P.B. Littlewood, B. Deveaud, and L.S. Dang, Nature 443, 409–414 (2006).

5 R. Balili, V. Hartwell, D. Snoke, L. Pfeiffer, and K. West, Science (New York, N.Y.) 316, 1007–10 (2007).

6 H. Deng, D. Press, S. Götzinger, G. Solomon, R. Hey, K. Ploog, and Y. Yamamoto, Physical Review Letters 97, 146402 (2006).

7 H. Deng, G. Weihs, C. Santori, J. Bloch, and Y. Yamamoto, Science 298, 199–202 (2002).

8 M. Assmann, J.-S. Tempel, F. Veit, M. Bayer, A. Rahimi-Iman, A. Löffler, S. Höfling, S. Reitzenstein, L. Worschech, and A. Forchel, Proceedings of the National Academy of Sciences of the United States of America 108, 1804–9 (2011).

9 J. Kasprzak, M. Richard, A. Baas, B. Deveaud, R. André, J.-P. Poizat, and L. Dang, Physical Review Letters 100, 067402 (2008).

10 P. Hohenberg, Physical Review 158, 383–386 (1967).

11 O. El Daïf, A. Baas, T. Guillet, J.-P. Brantut, R.I. Kaitouni, J.L. Staehli, F. Morier-Genoud, and B. Deveaud, Applied Physics Letters 88, 061105 (2006).

12 N.Y. Kim, C.-W. Lai, S. Utsunomiya, G. Roumpos, M. Fraser, H. Deng, T. Byrnes, P. Recher, N. Kumada, T. Fujisawa, and Y. Yamamoto, Physica Status Solidi B 245, 10 (2008).

13 T. Gutbrod, M. Bayer, A. Forchel, and J.P. Reithmaier, Physical Review B 57, 9950–9956 (1998).

14 D. Bajoni, E. Semenova, A. Lemaître, S. Bouchoule, E. Wertz, P. Senellart, and J. Bloch, Physical Review B 77, 113303 (2008).

15 S.I. Tsintzos, P.G. Savvidis, G. Deligeorgis, Z. Hatzopoulos, and N.T. Pelekanos, Applied Physics Letters 94, 071109 (2009).

16 A.A. Khalifa, A.P.D. Love, D.N. Krizhanovskii, M. Skolnick, and J. Roberts, Applied Physics Letters 92, 061107 (2008).

17 D. Bajoni, Journal of Physics D: Applied Physics 45, 313001 (2012).

18 A. Verger, C. Ciuti, and I. Carusotto, Physical Review B 73, 193306 (2006).

19 N. Na and Y. Yamamoto, New Journal of Physics 12, 123001 (2010).

K. Winkler et al., page 10

20 A. Rahimi-Iman, C. Schneider, J. Fischer, S. Holzinger, M. Amthor, S. Höfling, S. Reitzenstein, L. Worschech, M. Kamp, and A. Forchel, Physical Review B 84, 165325 (2011).

21 C. Schneider, T. Heindel, A. Huggenberger, T.A. Niederstrasser, S. Reitzenstein, A. Forchel, S. Höfling, and M. Kamp, Applied Physics Letters 100, 091108 (2012).

22 N.Y. Kim, K. Kusudo, C. Wu, N. Masumoto, A. Löffler, S. Höfling, N. Kumada, L. Worschech, A. Forchel, and Y. Yamamoto, Nature Physics 7, 681–686 (2011).

23 I. Carusotto, D. Gerace, H. Tureci, S. De Liberato, C. Ciuti, and A. Imamoglu, Physical Review Letters 103, 033601 (2009).

24 P. Lugan, D. Sarchi, and V. Savona, Physica Status Solidi (C) 3, 2428–2431 (2006).

25 A.L. Bradley, J.P. Doran, J. Hegarty, R.P. Stanley, U. Oesterle, R. Houdré, and M. Ilegems, Journal of Luminescence 85, 261–270 (2000).

Figure captions

Fig. 1: (a) Schematic sketch of the sample design. The doping levels are indicated by white

(intrinsic) to black (high doping) color coding. The QW is drawn in red color. (b) Fitted

ground mode emission of an electrically driven 10 µm trap to obtain the quality factor of

6150. (c) Scanning electron microscope (SEM) image of a sample processed for electrical

injection. The image shows pillars of 30 µm diameter with gold contact rings around the

centered traps.

Fig. 2: Characterization of the planar polaritons in means of 2D photoluminescence

measurements. (a) Anticrossing of upper and lower polariton extracted from position resolved

investigations. (b) Momentum resolved dispersion of the planar polariton at a negative

detuning of -1.77 meV. (c) Dispersion of the trapped polariton recorded at a detuning of -0.9

meV.

K. Winkler et al., page 11

Fig. 3: Emission pattern from a trap with diameter of 10µm. Comparison of (a)

photoluminescence and (c) electroluminescence measurements. The false-color images show

quantized trap states as well as continuous branches from planar background. The raw

maxima are fitted in (b) (PL) and (d) (EL) by the theoretical polariton dispersions.

Fig. 4: Power-dependent EL of a 2µm trap. At low injection currents, the Rabi splitting is

obvious in the planar background emission (a), while the dispersion becomes photon-like

(indicated by black line) at high pump rates (b). (c) Power-dependent integrated intensity and

energy shift of the ground mode.

PL Measurement Fitted Curves Bare Dispersions

1.402

1.404

1.406

1.408

1.410(c) 10μm Trap(b) Planar

Ene

rgy

(eV

)

(a)

-5 -4 -3 -2 -1 0 1 2 3 4 5Detuning EC-EX (meV)

ER

EX

EC (2D)

EUP (2D)

ELP (2D)

-2 -1 0 1 2

k|| (μm-1)

LP2D

UP2D

-2 -1 0 1 21.400

1.402

1.404

1.406

1.408

1.410

1.412

1.414

UP2D

UPTrap

LPTrap

Ene

rgy

(eV

)

1.398

1.400

1.402

1.404

1.406

1.408

1.410

1.412

1.414

1.416

1.418 (c) EL of 10μm Trap

Ene

rgy

(eV

)

-1 0 1

(d)

-2 -1 0 1 2

Fitted Curves Raw Maxima Bare Dispersions

ELP (2D)

k|| (μm-1)

EC (2D)

ELP (Trap)

EC (Trap)

EX

EUP (2D)

-2 -1 0 1 21.398

1.400

1.402

1.404

1.406

1.408

1.410

1.412

1.414

1.416

1.418 (a) PL of 10μm Trap

k|| (μm-1)

Ene

rgy

(eV

)

-2 -1 0 1 2

Fitted Curves Raw Maxima Bare Dispersions

EUP (2D)

EC (2D)

EC (Trap)

ELP (Trap)

(b)

EX

10.1

1

10

Intensity ground mode 2µm

IntegratedIntensity(a.u.)

Current (mA)

EC (Trap)(c)

1.4007

1.4008

1.4009

1.4010

1.4011

1.4012

Energy(eV)

Energy ground mode 2µm

1.400

1.405

1.410

1.415

(a) 0.4 mAUP2D

LPTrap

LP2D

-1 0 1

1.400

1.405

1.410

1.415

(b) 4.2 mA

C2D

k|| (µm-1)

Energy(eV)

LPTrap

Copyright (2013) American Institute of Physics. This article may be downloaded for personal

use only. Any other use requires prior permission of the author and the American Institute of

Physics.

This article appeared in Applied Physics Letters and may be found at

http://dx.doi.org/10.1063/1.4777564.